Niels Pinkwart

Predicting MOOC Dropout over Weeks Using Machine Learning Methods

With high dropout rates as observed in many current larger-scale online courses, mechanisms that are able to predict student dropout become increasingly important. While this problem is partially solved for students that are active in online forums, this is not yet the case for the more general student population. In this paper, we present an approach that works on click-stream data. Among other features, the machine learning algorithm takes the weekly history of student data into account and thus is able to notice changes in student behavior over time. In the later phases of a course (i.e., once such history data is available), this approach is able to predict dropout significantly better than baseline methods.

Toward legal argument instruction with graph grammars and collaborative filtering techniques

This paper presents an approach for intelligent tutoring in the field of legal argumentation. In this approach, students study transcripts of US Supreme Court oral argument and create a graphical representation of argument flow as tests offered by attorneys being challenged by hypotheticals posed by Justices. The proposed system, which is based on the collaborative modeling framework Cool Modes, is capable of detecting three types of weaknesses in arguments; when it does, it presents the student with a self explanation prompt. This kind of feedback seems more appropriate than the “strong connective feedback” typically offered by model-tracing or constraint-based tutors. Structural and context weaknesses in arguments are handled by graph grammars, and the critical problem of detecting and dealing with content weaknesses in student contributions is addressed through a collaborative filtering approach, thereby avoiding the critical problem of natural language processing in legal argumentation. An early version of the system was pilot tested with two students.

LASAD: Flexible representations for computer-based collaborative argumentation

Teaching argumentation is challenging, and the factors of how to effectively support the acquisition of argumentation skills through technology are not fully explored yet. One of the key reasons for that is the lack of comparability between studies. In this article, we describe LASAD, a collaborative argumentation framework that can be flexibly parameterized. We illustrate the flexibility of the framework with respect to visualization, structural definitions and kind of cooperation. Using this framework, this paper presents an evaluation of the impact of using an argumentation system with different argument representations and with collaborative vs. individual use on the outcomes of scientific argumentation. We investigate which combinations of these factors produces the best results concerning argument production and learning outcomes. The results of this controlled lab study with 36 participants showed that the use of simple representational formats is superior compared to highly structured ones. Even though the latter encouraged the provision of additional non-given material, the former is less error-prone. A hypothesized structural guidance provided by more complex formats could not be confirmed. With respect to collaboration, the results highlight that arguing in groups lead to more cluttered argumentation maps, including a higher amount of duplicate elements. An expected peer-reviewing between group members did not occur. Yet, groups also tended to include more points-of-view in their arguments, leading to more elaborated argument maps.

Using Collaborative Filtering Algorithms as eLearning Tools

Collaborative information filtering techniques play a key role in many Web 2.0 applications. While they are currently mainly used for business purposes such as product recommendation, collaborative filtering also has potential for usage in eLearning applications. The quality of a student provided solution can be heuristically determined by peers who review the solution, thus effectively disburdening the workload of tutors. This paper presents a collaborative filtering approach which is specifically designed for eLearning applications. A controlled lab study with the system confirmed that the underlying algorithm is suitable as a diagnostic tool: The system-generated quality heuristic correlated highly with an expert-provided manual grading of the student solutions. This was true independent of whether the students provided fine-grained or coarsegrained evaluations of peer solutions, and independent of the task type that the students worked on. Further, the system required only few peer evaluations in order to achieve an acceptable prediction quality.

Learning by diagramming Supreme Court oral arguments

This paper describes an intelligent tutoring system, LARGO, that helps students learn skills of legal reasoning with hypotheticals by analyzing oral arguments before the US Supreme Court. The skills involve proposing a rule-like test for deciding a case, posing hypotheticals to challenge the rule, and responding by analogizing or distinguishing the hypotheticals and/or modifying the proposed test. Students diagram arguments in a special-purpose graphical language and receive feedback in the form of reflection questions.

A Review of AI-Supported Tutoring Approaches for Learning Programming

In this paper, we review tutoring approaches of computer-supported systems for learning programming. From the survey we have learned three lessons. First, various AI-supported tutoring approaches have been developed and most existing systems use a feedback-based tutoring approach for supporting students. Second, the AI techniques deployed to support feedback-based tutoring approaches are able to identify the student’s intention, i.e. the solution strategy implemented in the student solution. Third, most reviewed tutoring approaches only support individual learning. In order to fill this research gap, we propose an approach to pair learning which supports two students who solve a programming problem face-to-face.

CoChemEx: Supporting Conceptual Chemistry Learning Via Computer-Mediated Collaboration Scripts

Chemistry students, like students in other disciplines, often learn to solve problems by applying well-practiced procedures. Such an approach, however, may hinder conceptual understanding. We propose to promote conceptual learning by having pairs of students collaborate on problems in a virtual laboratory (VLab), assisted by a computer-mediated collaboration scriptthat guides the students through the stages of scientific experimentation and adaptsto their needs for support. We used the results from a small-scale study comparing how singles and dyads solve chemistry problems with the VLab with and without scripts to develop a scripted collaborative experimentation environment. A subsequent small-scale study compared an adaptive and a non-adaptive version of the system. Qualitative data analyses revealed a tendency for the dyads in the adaptive feedback condition to improve their collaboration and be more motivated than the non-adaptive dyads. In this paper, we present our research framework and report on preliminary results from the two small-scale studies.

A Discrete Particle Swarm Optimization Approach to Compose Heterogeneous Learning Groups

Collaborative learning is an educational strategy which is popularly used in project-based courses in schools and colleges. The diversity of group members is frequently considered to be a crucial criterion that can promote intensive intra-group interaction and successful learning outcomes. Yet, when the number of students is up to several hundreds, it is challenging for instructors to look for an optimal group formation considering maximal diversity of students in every group. To address this problem, this paper presents a discrete particle swarm optimization approach to compose heterogeneous learning groups. We carried out simulations based on optimizing the heterogeneity of gender and personality type. The experimental results show that the proposed approach is an effective and stable method that can support instructors to compose heterogeneous collaborative learning groups.

Example-based feedback provision using structured solution spaces

Intelligent tutoring systems (ITSs) typically rely on a formalised model of the underlying domain knowledge in order to provide feedback to learners adaptively to their needs. This approach implies two general drawbacks: the formalisation of a domain-specific model usually requires a huge effort, and in some domains it is not possible at all. In this paper, we propose feedback provision strategies in absence of a formalised domain model, motivated by example-based learning approaches. We demonstrate the feasibility and effectiveness of these strategies in several studies with experts and students. We discuss how, in a set of solutions, appropriate examples can be automatically identified and assigned to given student solutions via machine learning techniques in conjunction with an underlying dissimilarity metric. The plausibility of such an automatic selection is evaluated in an expert survey, while possible choices for domain-agnostic dissimilarity measures are tested in the context of real solution sets of Java programs. The quantitative evidence suggests that the proposed feedback strategies and automatic example assignment are viable in principle, further user studies in large-scale learning environments being the subject of future research.

Autonomous agents in organized localities regulated by institutions

This paper proposes a new metaphor for constructing systems of systems: Autonomous Agents in Organized Localities (AAOL). An agent-based approach is used for modeling structure and behavior of complex systems that consist of (semi-)autonomous systems, where goals, resources, capabilities are described locally while a need for superordinated ”global” regulation exists. The notion of organized localities is used to describe spatially or logically constrained spheres of influence of regulation bodies. Agents inhabit — and can move across — localities; regulation rules are modeled via computational norms and enforced by electronic institutions. A key objective of our work is to explore and advance applicability of AAOL to constructing mechatronic systems with (at least soft) real-time constraints. We describe requirements for modeling systems of systems, and outline the key pillars of AAOL: a conceptual architecture and a metamodel providing the basic constructs for describing AAOL-type systems. A case study of a decentrally organized airport transportation infrastructure illustrates the concepts and the feasibility of AAOL-based systems of systems design.